I watched the first 2 lectures of the Part 2 course, and wanted a hands on experience so I went to Google Colab to install fastai v0.7.0 (and all dependencies and specific versions). The library worked nicely until I reached a line of code that said:
grid_sizes = V(np.array([1/anc_grid]), requires_grad=False).unsqueeze(1)
which I reckon is the first time the V() method is called. Whenever i ran this cell in the notebook, the cell would keep on running until all the RAM was utilized and the notebook crashed and restarted. This only occurs when i am using the hardware acceleration (GPU) in the notebook. Without hardware acceleration, the cell runs fine and the model is trained.
For some testing, is just tried calling the method as:
a = V(1)
and this cell also hangs/runs indefinitely and eventually runs out of ram and crashes.
I know that in a few months the 2nd part based on V1 fastai would be uploaded, but I was hoping if there’s a quick solution to this, please let me know.